Non-Invasive and cost effective monitoring of energy consumption patterns for electrical equipment.

Electricity use in many buildings is increasing, sometimes such that new buildings built to modern standards can have greater energy usage than their predecessors. A major cause is increased use of appliances such as computer equipment, lighting, and electronics. Increasing standby power consumption and more appliances being left running unnecessarily, compounds these effects. Monitoring equipment should provide insights into these issues, but the few systems currently available are either expensive, invasive, or both. This paper shows a case study where temperature monitoring was used to reduce industrial plant usage, and shows a method of electrical appliance monitoring, using very low cost sensors. Results show that the method can effectively reveal equipment usage patterns. Nomenclature b filter coefficients (unitless) Ts standby temperature (C) Gt temperature gradient (unitless) Z z transform operator (unitless) H filter transfer function (unitless) CT current transformer N filter order (unitless) LDR light dependent resistor Ta ambient temperature (C) OCR optical character recognition To max. operating temperature (C) OMR optical meter reader Tp instantaneous temperature (C) Introduction Buildings account for almost 50% of C02 emissons in the UK [1]. As part of the Carbon Reduction in Buildings Project, the authors investigated new ways of analysing half-hourly data at the appliance level [2], noting a scarcity of suitable appliance monitoring systems. While plug in power meters provide a basic audit of appliances, usage patterns are not established, since the power meters do not log. Energy use questionnaires have worked well, but discrepancies may exist between perception of energy use and reality [3]. Non-invasive sensing The key concept described in this paper is the use of non-invasive temperature sensing for appliance usage patterns. While temperature logging of plant has been carried out in the past to determine machinery ‘health’, for example lubricating oil temperatures in HVAC plant, we propose a more wide ranging strategy of temperature monitoring for tens or even hundreds of electrical appliances within a building, in order to build a complete picture of electrical energy consumption. We also propose a method of data post processing which produces easily identifiable duty cycles as opposed to raw temperature data. As a result, the data is readily processable by software, which becomes crucial for large audits, for example of all IT hardware within an office block. The method is low cost. A further advantage of case temperature monitoring is that personnel carrying out energy audits do not need to remove equipment covers, or work with high voltage cabling. Whole system (building) monitoring Energy meters can be read manually for patterns of overall appliance use, which is entirely noninvasive and usually very reliable, provided the readings actually occur. But this is not practical for high frequency (<hourly) readings. For automatic logging, some energy meters may give a contact closure, or increasingly commonly, a solid state output. The connection usually takes the form of an RJ10 connector, similar to a UK telephone plug [4]. Where a pulse output meter is fitted, logging is theoretically easy. However, energy suppliers’ permission to utilise UK energy meter electrical pulses is required. Noting the complicated structure of UK energy supply, dialogue with energy companies is difficult, therefore non-contact sensing data collection being magnetic or optical, or recently, OCR based. Where possible, the simplest option is to utilise a meter which produces a magnetic pulse. The meter casing remains un-breached, and therefore maintains compliance with ingress protection standards for electricity meters. An OMR [5] uses a send and receive pair of photodiodes, to view the passage of analogue dial pointers, e.g. from the revolving disk of an electricity meter. A pulse logger stores in memory the number of meter pulses for a given time period. The authors have developed a solution for pulse logging where electricity meters feature a flashing LED. A NORPS-12 type LDR, is able to provide an output which directly drives a small pulse logger [6]. This provides a compact solution for pulse counting over a programmable interval, with a storage capacity for half hourly metering of around 400 days. CTs may be used to detect power consumption, the proviso being that a clear phase is required for sensing (multicore Live Neutral and Earth cable, for example, would not be suitable). Another disadvantage of CT based devices is poor sensitivity for low loads. Individual appliance monitoring Appliance monitoring systems differ from plug in power meters in that logging is possible. The Watteco, relied on deciphering circuit behaviour [7] although current trends suggest that powerline communications [8] may prove more reliable. Such systems can produce very high quality data, but may not be required for a basic energy audit. Another method of measuring appliance usage may involve sense the magnetic field of a motor or flourescent ballast [9]. It is fundamental that in the majority of cases, an electrical appliance will become warm through normal use. Main causes of this may be due to electrical inefficiences (such as heat generated by electronics), magnetic circuits (eddy current heating of transformer cores), deliberate heating (e.g. a cathode ray tube or heating element), or process heat (e.g. a washing machine or airconditioner). Measurement of case temperatures provides a low cost method of usage pattern monitoring. A summary of the available technologies for individual appliance monitoring is provided below in Table 1. Table 1 – Summary of technologies for appliance monitoring Technology Approx. Base cost for hardware, software. (Euro) Approx. Cost (Euro) per appliance Notes Temperature loggers 50 25 Imprecise, post processing needed Power meters (plug in) 0 25 Not capable of profile logging but capable of measuring integrated consumption Power meters (benchtop) 25 300 Bulky – lab apparatus used for calibration but capable of logging and very accurate Current clamps 40 150 Manual data collection required, certain current transformers may not operate at low currents Magnetic field sensor 25 60 Needs to be positioned close to motor or flourescent light ballast Individual appliance monitoring 1000 120 Extremely accurate and precise, automatic data uploads include integrated consumption Case Study – non-invasive industrial forensic monitoring using case temperatures The East Midlands New Technology Initiative (NTI) in the UK, is backed by leading colleges and universities in the region. NTI provides grants to small and medium sized businesses to invest in state of the art technology and gain advanced technology skills through NTI approved courses. As part of this work, an electronics factory was visited in order to carry out forensic plant monitoring. The factory is an SME, employing around 40 staff (Figure 1, figure 2), and is occupied from Monday to Friday for production, although some administration occurs on Saturdays. Using an Eltek radio datalogging system, two items of plant were tested, a flow soldering machine (figure 3), and an air compressor located outside, shown in figure 4, probes clipped to hot air exhausts for each item. Data was logged at 5 minute intervals for 14 days. A Willow optical meter reader was used to monitor electricity consumption from the main site electricity meter. As can be seen from figure 5, the temperature of the flow soldering machine (the largest item of plant at the factory) approximately correlates with overall electrical load. (Th2 is ambient indoor temperature). During the investigation, this machine was the subject of an energy management strategy, and was timed to preheat shortly before the start of each shift. A 24 hour time switch supplied the compressor to switch off at night. As can be seen from the plot, compressor temperature clearly indicates weekend running, where the compressor ran due to system leakage. As well as tackling air system leaks, the compressor timer was changed to prevent weekend operation. This shows an example of a simple application of temperature monitoring for identification of plant usage, with energy savings. Figure 1 – Factory Exterior Figure 2 – Factory interior Figure 3 – Flow Soldering Machine Exhaust Figure 4 – Compressor Hot Air Exhaust Figure 5 Factory plant case temperatures and overall factory electricity consumption Data post processing Thresholding of temperature data, either by sight or using automated methods, is effective in producing a rough estimate of appliance usage, but cannot compete with direct power measurement for precise appliance use estimation. Subtraction of ambient temperatures may assist in improving the reliability of thresholding, but its use is limited, since appliances with a high thermal mass may never cool enough when switched off to cross the threshold. An alternative approach which should produce more accurate data is to look at the heating/cooling cycle. Case temperature of an electrical appliance typically follows a set pattern. Referring to figure 6, when the appliance is at rest, case temperature equals ambient (A). When power is applied, case temperature rises (B) until a maximum is reached. Unless varying loads are present (e.g. for a washing machine, computer), operating temperature (To) is typically steady. When power is disconnected, case temperature will drop back to ambient, the slope (D) usually being less than that of B. The case temperature will then return to ambient (E). If an appliance is left on standby, the small power drain will usually affect case temperature, such that the standby temperature (Ts) will be several degrees above ambient. In reality, some post processing must be applied for precise, automatic analysis. The intention is to produce a stream of binary data, reflecting appli